I've seen many UI rollouts that carefully track user happiness metrics. All UI changes have in common that user happiness goes down for a short time after the change -- but for good changes, it goes back up and to an even higher level than before. So it's expected that right after a change people are unhappy, which is where Nielsen's (sensible) suggestion comes from.

The problem is of course that UX designers get so used to dismissing that initial criticism that they sometimes dismiss valid criticism that leads to long-term user unhappiness as well.